Generated by Codex with GPT-5

This summary covers The Economist’s June 13th, 2026 Business article listed in the contents as Rise of the robo-couriers and published under the headline Pavement pioneers.

Delivery robots have long sounded like a city-of-the-future novelty: small machines rolling along pavements, carrying groceries or pizza while pedestrians decide whether to be impressed, annoyed or amused. The article argues that the novelty is becoming a business proposition. In Milton Keynes, a British city better known for roundabouts than futuristic technology, Starship Technologies has built one of the largest markets for autonomous grocery delivery. Its claim is simple and commercially important: small pavement robots can deliver goods more cheaply than human drivers.

Why small robots are easier than robotaxis

Starship’s machines face many of the same problems as self-driving cars. Each robot needs sensors that can read a messy outdoor environment in rain or shine. Those sensors feed an artificial-intelligence system that chooses routes, avoids obstacles and keeps moving even when the connection to a data centre drops. The robots are not just remote-controlled carts; they are small autonomous vehicles that must handle the everyday disorder of pavements.

Yet the comparison with robotaxis also shows why delivery robots may arrive sooner. A six-wheeled courier weighing 35kg and moving at about walking speed poses far less danger than a two-tonne car on a motorway. Its cargo is also more forgiving. A slightly bumpy ride may be irritating for a passenger, but it will not much trouble a pizza.

That lower risk matters because autonomy is often constrained less by clever demos than by edge cases, liability and public trust. The pavement robot has an easier bargain to offer: accept some clutter on the sidewalk in exchange for cheaper local delivery and fewer short car trips.

The hard work is in the machine itself

The article is careful not to present the technology as effortless. Robotaxi companies can often adapt existing vehicles from carmakers. Starship does not have that luxury. It has had to design and refine its own hardware over several generations, making the robots sturdier, easier to repair and cheaper to operate.

Some improvements are prosaic but decisive. The latest robots charge wirelessly, which speeds up turnaround and reduces wear on physical charging ports. Better repairability lowers downtime. Better routing and autonomy reduce the need for human intervention. None of these changes is as dramatic as a breakthrough in artificial intelligence, but together they move the economics from experiment toward infrastructure.

Starship says each delivery now costs significantly less than paying a person. Its target is less than GBP1, or about \$1.34, per delivery. It is not quite there yet, according to co-founder Ahti Heinla, but it is close enough for the claim to matter. Last-mile delivery is a low-margin, operationally messy business. If robots can reliably shave costs at scale, they do not need to be glamorous to be disruptive.

Scale changes the problem

The company’s progress is visible in the numbers. In 2018 Starship had 127 robots that travelled 116,000 kilometres over the year. By 2025 it had 2,414 robots covering 5.2m kilometres. Over that period it cut human interventions per kilometre by seven-eighths.

That is the right direction, but scale creates its own burden. A rare failure matters more when thousands of robots are operating across a city. A robot stuck in the middle of a street is not just a technical defect; it is a public nuisance and a reputational risk. Starship’s answer includes a simple backup computer whose only job is to get a stranded robot to the other side of the road.

The social complications may be as important as the engineering ones. Pavement robots will irritate some pedestrians. They may threaten delivery work. They may also provoke unexpected affection. In Finland, Starship’s largest market, its supermarket partner has had to ask helpful passers-by not to lift stuck robots out of snowdrifts, partly to prevent injuries.

The article’s larger point is that autonomous delivery is likely to enter city life quietly, through mundane errands rather than science-fiction drama. The first successful robots may not be humanoid assistants or fast-moving robotaxis. They may be slow, boxy, repairable machines carrying groceries across town, cheap enough to make the economics work and harmless enough for people to tolerate.